University of Twente Student Theses


Analyzing the predictive capacity of a physically-based debris flow model: a case study for debris flow in the post-2008 Wenchuan earthquake area.

Xiao, Yi (2020) Analyzing the predictive capacity of a physically-based debris flow model: a case study for debris flow in the post-2008 Wenchuan earthquake area.

Link to full-text:
(only accessible for UT students and staff)
Abstract:Post-seismic debris flows are one of the most destructive post-earthquake secondary hazards. The threats posed by them can last decades. One of the most effective methods of risk mitigation is debris flow early warning. Physically-based models are widely used to reproduce historical debris flows, and they performed excellently. It is believed that the physically-based models are capable of high feasibility on debris flow prediction. And the transferability of such models in space and time has been widely assumed that specific calibration is required specific calibration in general. To investigate the predictive capacity of physically-based models, an integrated debris flow physically-based model (OpenLISEM Hazard) was applied to the earthquake-stricken area in Wenchuan, China. In order to evaluate the spatial transferability, the model was calibrated for three catchments separately based on debris flow inventory. After that, cross-validations of calibrated models were operated between the three watersheds. Results show that the model performed well in simulating occurrence and scale of debris flows, the prediction of the precise arrival time of debris flow is a bit problematic. Cohen’s kappa of spatial transferability regarding accuracy volume and time are 0.02 and 0.38 separately. The volume accuracy is affected by the transferred calibrated parameters in space, while the prediction of the exact arrival time is impacted less. Besides, the sensitivity of parameters is not only model-specific but also catchment-specific. To evaluate the temporal transferability, four variables that change over time were considered, and the calibrated models were run with different rainfall events from 2010 to 2013. The temporal transferability was investigated from three aspects. Two aspects are to test whether the model underpredicts the occurrence of debris flows and whether the calibrated models can produce the exact arrival time of debris flows over time. For these purposes, the calibrated models were run with the rainfalls that triggered debris flows. The results state that the model does not have the problem of low prediction of occurrence. However, in two watersheds, the arrival times were given at similar model-running times every year. Another aspect is to investigate whether the model overpredicts the occurrence of debris flows. In this case, the calibrated models were run with the rainfalls that did not trigger debris flows. The results show that the model is a little overproduce processes of debris flows. This study, therefore, suggests that the calibrated parameters of OpenLISEM Hazard can be spatially and temporally transferred between similar catchments and within at least five years in the earthquake-stricken area to predict the scale and occurrence of debris flows. The predictive capacity of the exact time is required to be improved. The integrated physically-based model is worth being expected to be applied for early warning in the broad region over time.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page